The Orc Quest Example

Resource-based planning systems have recently begun to mushroom | an indication that AI planning systems have nally started to tackle real-world problems. However, resources tend to be neglected. Conventional plan length is still the primary optimization goal, resources being of only secondary importance. This paper illustrates the problem and gives an overview of the Excalibur agent's planning system, which overcomes these deeciencies by applying an extended constraint programming framework to planning. It allows the search space to be explored without the need to focus on plan length, makes it possible to optimize other criteria like resource-related properties and promotes the inclusion of domain-dependent knowledge to guide and accelerate the search. Imagine you are a computer-guided Orc in a computer game 1. Your interests are somewhat diierent from those of your philistine Orc friends. To get them a little more acquainted with the ne arts, you decided to surprise them with a ballet performance. As a highlight, ve of those delicate humans are to play supporting roles. However, the uncivilized humans do not appreciate the charm of Orc-produced art and are extremely reluctant to participate. This makes it rather painful to catch a human and force him to take part in the performance, and much more painful to catch a whole group of them. On the other hand, going for a group saves time. This can be formalized as follows: 1 Sorry about the unusual setting but our application domain is computer games. For those who have not heard of Orcs before: \Orcs ... are the most famous of Tolkien's creatures. ... Orcs tend to be short, squat and bow-legged, with long arms, dark faces, squinty eyes, and long fangs. ... Orcs hate all things of beauty and love to kill and destroy." (Mattsson 1999) The Plan-Length Criterion Being an Orc, you lack suucient brain power to solve the problem. You therefore apply a conventional AI planning system and come up with the following plan: catch a group & catch a group, which yields a plan duration of 10 hours, an 8 on the pain scale and 6 performers. The plan attains your goal of catching at least 5 performers, but the other results look to you to be capable of improvement. You realize that you have forgotten to apply an optimization criterion. You thus repeat the planning process , applying a state-of-the-art resource-based planning system that you request to minimize …